HFBoards Data: Goaltender Goal Support (GFA, Goals For Average)
 Register FAQ Members List Search Today's Posts Mark Forums Read
 Mobile Hockey's Future Become a Sponsor Site Rules Support Forum vBookie Page 2
 By The Numbers Hockey Analytics... the Final Frontier. Explore strange new worlds, to seek out new algorithms, to boldly go where no one has gone before.

Data: Goaltender Goal Support (GFA, Goals For Average)

 07-18-2014, 01:00 PM #1 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 Data: Goaltender Goal Support (GFA, Goals For Average) Just added a new metric onto my site that I've always had a hard time finding. I call it GFA (Goals For Average) and I've tried to make it an analogue and direct comparable to GAA (Goals Against Average). It's how many goals of support a goaltender receives per game (per 60 minutes). All of the data come from my game logs data (and so currently only goes back as far as that). I removed empty net goals scored by the goaltender's team (as well as estimated empty net time). The one place where I'm not entirely happy with it is in games shared by a goaltender. Ideally (or perhaps not?) if Goaltender A was in net for the first 30 minutes (and his team scored 2 goals) and Goaltender B was in net for the last 30 minutes (and his team scored 0 goals) then Goaltender A would get 2 goals of support and Goaltender B would get 0 goals of support. However, in my game logs, I only have the game-level totals of support. I'm reasonably happy with this compromise, since ultimately, goal totals only matter at the end of the game (*) and in the grand scheme of things, it doesn't really matter when they were scored during the game. That's somewhat correct and somewhat flawed, of course. (*) You could also validly argue that goal totals also only matter at the end of the game in relation to the other team's goal totals. That's already measured, although not perfectly, in win-loss records. Anyhow, these are now on the individual goaltender pages, for instance: http://hockeygoalies.org/bio/brodeurm.html http://hockeygoalies.org/bio/roy.html http://hockeygoalies.org/bio/hasek.html I'd like to put GFA right next to GAA, since that's the obvious comparable, although GAA is widely accepted while GFA is new so I feel weird doing that just yet. And on the team-level pages, for instance: http://www.hockeygoalies.org/bio/nhl/newjersey.html The team-level pages are perhaps more interesting, since you can compare goaltender (Schneider) vs. goaltender (Brodeur). Anyhow, enjoy - and comments are welcome (encouraged).
 07-18-2014, 01:14 PM #2 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 One interesting thing I've noticed - Brodeur's win/loss record this past year (relative to Schneider's) is largely related to the massive difference in goal support received by Brodeur (43% more). Some have argued that (somehow) Brodeur's style of play is the direct cause of this. If that's the case, then I would expect to see it even more when Brodeur was in his prime years, and I'm not seeing a consistent pattern of results in the data.
 07-18-2014, 11:01 PM #3 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 Sitting here staring at Grant Fuhr: http://hockeygoalies.org/bio/fuhr.html People always talk about his (relatively) high goals-against averages. The comparable goals-for averages put things in proper perspective. For instance, Fuhr and the Oilers allowed 3.93 goals per game in 1985-86. How did he go 29-8? It's easy when your team is putting up 5.61 (!) every sixty minutes.
07-19-2014, 07:31 AM
#4
hatterson
Global Moderator

Join Date: Apr 2010
Location: North Tonawanda, NY
Country:
Posts: 12,583
vCash: 50
Quote:
 Originally Posted by Chalupa Batman The one place where I'm not entirely happy with it is in games shared by a goaltender. Ideally (or perhaps not?) if Goaltender A was in net for the first 30 minutes (and his team scored 2 goals) and Goaltender B was in net for the last 30 minutes (and his team scored 0 goals) then Goaltender A would get 2 goals of support and Goaltender B would get 0 goals of support. However, in my game logs, I only have the game-level totals of support.
I'll pop in for the "or perhaps not" crowd.

When a goalie starts a game he obviously starts it at 0-0. Boiled down, his job is ultimately to allow fewer goals than the goalie across the rink from him. However, when a goalie comes into a game midway through it is rarely 0-0. Unless he's coming in due to injury, his team is generally down multiple goals.

In a way, at the moment of entering the game he's already received -2 or -3 goals of support. His job isn't to simply allow fewer goals, it's to allow 3 or 4 fewer goals.

In the rarer case where a goalie comes in with a lead, he's essentially already received a goal or three of support. Simply because they weren't scored when he was in goal, doesn't mean he doesn't benefit from them.

Granted, I'm not sure how much this would actually affect any analysis, especially for regular starters and they wouldn't come on in relief near as much, however it might be interested to see a GFA for starts only, excluding games where the goalie didn't start the game.

__________________
Come join us on the By The Numbers forum. Take a look at our introduction post if you're new. If you have any questions, feel free to PM me.

 07-19-2014, 09:24 AM #5 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 That's a really good point - Tom Barrasso allowing 4 goals (in 30 minutes) to start the game is different than Barrasso allowing 4 goals (in 30 minutes) if he comes in with an 8-1 lead. Ideally, I'd present both (although my "both" philosophy is the primary contributor to the bio pages getting too wide for a single screen ). I like the GFA in starts idea - that's an easier calculation even, since it's just (totals goals for) / (starts).
07-19-2014, 08:20 PM
#6
Caeldan
Moderator
Whippet Whisperer

Join Date: Jun 2008
Country:
Posts: 9,740
vCash: 50
Quote:
 Originally Posted by Chalupa Batman That's a really good point - Tom Barrasso allowing 4 goals (in 30 minutes) to start the game is different than Barrasso allowing 4 goals (in 30 minutes) if he comes in with an 8-1 lead. Ideally, I'd present both (although my "both" philosophy is the primary contributor to the bio pages getting too wide for a single screen ). I like the GFA in starts idea - that's an easier calculation even, since it's just (totals goals for) / (starts).
I think the 'ideal' way to do it would be GFA (calculated like GAA) but only for game starts? This way you remove score effects due to influences from the other goaltender? Though, how often do goalies get pulled and would it really make much difference in the long run?

As an aside... would it be difficult on team pages to flag 'rookie' goalies for seasons, be curious to see if teams tend to have their GF go down for rookie goalies (that's the general wisdom I believe, since they tend to play 'tighter' hockey since they don't expect the kid to come through and bail them out on bad pinches etc).

 07-19-2014, 09:35 PM #7 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 The rookie flags could be done, although I'd probably have to do it manually (set up a table in my database with Calder-eligible seasons, and join it with the main file before exporting). I do have almost something that could be a good proxy - I have age on the left side of each goaltender's stat line. So for instance: http://hockeygoalies.org/bio/nhl/ottawa.html The Senators scored 25% more goals last year for the 32-year-old Anderson than they did for the 22-year-old Lehner, which pretty much explains their W/L differential.
 07-23-2014, 01:22 PM #8 BraveCanadian Registered User     Join Date: Jun 2010 Country: Posts: 9,529 vCash: 500 I think it is nice to have even a reasonable barometer of how the team is scoring for a particular goaltender. It really puts some of the goaltender's performances into context.
 07-23-2014, 02:41 PM #9 AfroThunder396 Lou's Secret Sauce     Join Date: Jan 2006 Location: Hamburg, NY Country: Posts: 24,107 vCash: 835 2014 New Jersey Devils: Brodeur - 41 GP, 2.76 GFA, 2.51 GAA Schneider - 45 GP, 1.93 GFA, 1.97 GAA Silly.
07-23-2014, 02:55 PM
#10
Doctor No
HFB Partner

Join Date: Oct 2005
Posts: 4,131
vCash: 50
Quote:
 Originally Posted by BraveCanadian I think it is nice to have even a reasonable barometer of how the team is scoring for a particular goaltender. It really puts some of the goaltender's performances into context.
Thanks! I realized that I needed it for my predictive model (or rather, the part that wasn't working well was related to future winning percentage, which isn't super important on its own but is correlated to future playing time (since teams certainly think that it's relevant).

Quote:
 Originally Posted by AfroThunder396 2014 New Jersey Devils: Brodeur - 41 GP, 2.76 GFA, 2.51 GAA Schneider - 45 GP, 1.93 GFA, 1.97 GAA Silly.
Agreed - one of my hopes was that the data might show a consistent pattern in Brodeur's GFA, suggesting that he is influencing the team's goal scoring. In his prime years, there *may* be something there, but the problem is that his backups played so few games that it's hard to compare.

In recent years (post-prime) there's no discernible pattern in favor of Brodeur year to year - that could be related to a number of things (in fact, my main conclusion is that in attempting to answer one question, I've introduced a half dozen new questions ).

Some posters (at least one) has suggested that the additional 0.83 goals of support that Brodeur got per game was entirely caused by him, which is a stretch (I'm willing to look at data that shows otherwise).

 04-12-2015, 02:44 PM #11 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 I've finished the 2014-15 regular season update for the NHL, so am posting some interesting numbers (all totals, including goal support, goaltender variation, hot/cold, are on my site including the links below). Most goal support per game (min. 1000 minutes played): Ben Bishop 3.17 Kari Lehtonen 3.09 Henrik Lundqvist 2.95 Jaroslav Halak 2.94 Michael Hutchinson 2.93 Least goal support per game (min. 1000 minutes played): Jhonas Enroth (BUF) 1.74 Anton Khudobin 1.74 Mike Smith 1.83 Michal Neuvirth 1.87 (BUF) Viktor Fasth 1.93 All totals are separated by team (so Dubnyk gets totals for the Coyotes and for the Wild separately). __________________ The Goaltender Home Page (http://hockeygoalies.org) - Preserving Goaltender History since 1994
04-14-2015, 07:49 PM
#12
Micklebot
Registered User

Join Date: Apr 2010
Posts: 9,389
vCash: 500
Quote:
 Originally Posted by Doctor No The rookie flags could be done, although I'd probably have to do it manually (set up a table in my database with Calder-eligible seasons, and join it with the main file before exporting). I do have almost something that could be a good proxy - I have age on the left side of each goaltender's stat line. So for instance: http://hockeygoalies.org/bio/nhl/ottawa.html The Senators scored 25% more goals last year for the 32-year-old Anderson than they did for the 22-year-old Lehner, which pretty much explains their W/L differential.
This is great stuff.

I'd point out that Anderson (and Brodeur in the other posters example) is a far better puck handler than his back up, and while the metric is support the goalie gets, part of that is from the support the goalie gives his offense. Simply stopping up a puck behind the net vs letting it go past and into the corner can be the difference between a quick transition and getting hemmed into your end. Goalies far to often are completely ignore in the context of the offense they help generate.

Another potential explanation for the gap in some goalies and their backups could be the skaters reluctance to take chances with a lesser goalie. Risk averse teams will score less than teams that take chances.

 04-14-2015, 08:17 PM #13 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 Thanks - I appreciate it! And I most definitely agree with your two potential causes; a third is luck (or whatever term people prefer). A fourth is strength of schedule - which I calculate separately as well. If you're playing Buffalo all year long, you're going to get a lot more goal support (and win a lot more games). Of course, my primary example is Mackenzie Skapski (who played Buffalo twice and no one else): http://hockeygoalies.org/bio/skapski.html And he got a total of four goals of (non-empty net) support. So it's possible that I'm full of it.
 06-01-2015, 09:44 AM #14 Doctor No HFB Partner     Join Date: Oct 2005 Posts: 4,131 vCash: 50 I've updated these on the website for 1975-76, 1976-77, and 1977-78. Will dig into interesting tidbits next.
06-02-2015, 08:39 PM
#15
eklunds source
Registered User

Join Date: Jul 2008
Location: Ed Snider's basement
Posts: 8,325
vCash: 500
Quote:
 Originally Posted by Micklebot I'd point out that Anderson (and Brodeur in the other posters example) is a far better puck handler than his back up, and while the metric is support the goalie gets, part of that is from the support the goalie gives his offense. Simply stopping up a puck behind the net vs letting it go past and into the corner can be the difference between a quick transition and getting hemmed into your end. Goalies far to often are completely ignore in the context of the offense they help generate.
Back in 2008, Brodeur is a Fraud evaluated Brodeur's impact on shot prevention (I'm making the leap that if you're preventing a shot, it's by gaining possession (otherwise aren't you just delaying a shot?), and the first step to offense is gaining possession):

Quote:
 In conclusion, there appears to be some evidence that Brodeur may prevent about a shot per game compared to average, based on comparing the results of the two backup goalies with the most games played, although the sample sizes are pretty small. Since many observers would consider Brodeur to be the best in the league at the skills that may contribute to preventing shots, I'd guess if we isolate his effect that could be considered pretty close to the maximum possible positive effect. From looking at the results of Dunham and others like him, it might be possible that the gap could be a little bigger on the other end (i.e. some goalies are substantially below average), but I think in general it is probably reasonable that goalie shots against results may vary within the range of about +/- 1 shot per game on average.
If the best puckhandling goaltender in his prime (before the trapezoid was "invented" at that) affected his team by preventing less than 82 shots in a season, I find it very hard to believe he could also, post-trapezoid and into his late 30s, help his team score an extra 0.83 goals per game (that's 68 goals over a full season).

on-ice sh%

When Brodeur was on the ice, the Devils shot 8.58% at 5v5 - they scored a goal every 11.7 shots.

With Schneider: 5.90%, meaning they scored a goal every 17 shots.

After normalizing for ice time, the Devils managed an extra 13 shots in ~1,827 minutes in front of Brodeur (26.4 shots/60 for MB, 25.96 for CS)... and allowed an extra 27 shots against.

If someone can come up with a logical suggestion how Brodeur helped his team convert shots into goals about 45% more often while having a negligible affect on his teams' shot totals for/against, I'll be happy to investigate their suggestion further.. Until then, I'm chalking it up as "on-ice shooting percentages over 40 games are wild and mostly meaningless".

Last edited by eklunds source: 06-02-2015 at 09:20 PM.

 06-02-2015, 09:13 PM #16 gqmixmaster Registered User     Join Date: Jun 2006 Posts: 2,360 vCash: 50 Id like to see if there is a correlation between gfa and a goalies average 1st period sv% given game strategy will generally be more defensive with a lead, I wonder how much the timing of goals affects gfa

Forum Jump